OpenAI and NASK, Poland's National Research Institute, have signed an agreement to explore joint research on AI safety and cybersecurity. George Osborne, who heads OpenAI for Countries, confirmed it in a LinkedIn post at the end of October 2026. On the Polish side the deal involves NASK's AI Safety Research Center, which serves as the Polish AI Safety Institute.
The topics named are evaluating increasingly capable AI systems, finding security vulnerabilities, and building safeguards against emerging threats. The agreement is exploratory: the areas of technical collaboration are still to be identified, and no specific joint project has been announced. The signing sets out a direction for cooperation rather than a named project.
“AI safety cannot be achieved in isolation. It is a collaborative effort, and only through active cooperation between public AI Safety Institutes and frontier AI developers can we ensure that AI becomes truly reliable and safe to use.”
Radosław Nielek, Director of NASK, in the announcement as printed by EdTech Innovation Hub · edtechinnovationhub.com ↗
NASK's AI Safety Research Center is the Polish counterpart to the safety and security institutes OpenAI already works with in the UK and the US, which gives the agreement a public-sector footing. The institute's own research already reaches into the territory the deal names, and its cybersecurity team has worked with OpenAI's models before.
The deal also builds on work already done. In July, CERT Polska, the team inside NASK, was given access to OpenAI's cyber-capable models through the Daybreak initiative. Using them it found six vulnerabilities in MikroTik's RouterOS, and MikroTik has since released fixes. According to OpenAI, some of the flaws could be chained together to take over devices without any authentication. The episode gave both sides a concrete case of outside researchers using OpenAI's models on real software.
OpenAI's stated reason is that it needs expertise from outside to challenge its own assumptions, test its systems and strengthen safeguards. Osborne said NASK's experience of protecting public services helps ground international safety research in what institutions face in practice.
About NASK
NASK is Poland's National Research Institute, working in cybersecurity, artificial intelligence and cloud computing through three research divisions and 14 specialised departments. Its AI Safety Research Center serves as the Polish Artificial Intelligence Safety Institute, and CERT Polska sits inside the institute. Its research covers detecting AI-generated synthetic content, malicious use of AI and multilingual security challenges.
Beyond research, NASK supports the public sector. It connects schools to the internet, builds secure document systems for government and trains public officials in cybersecurity.
OpenAI is also pushing for international technical standards for frontier AI, covering shared evaluation, risk assessment, safeguard testing and incident reporting. The company already works with the UK AI Security Institute and the US CAISSI, and NASK now joins that circle of public bodies.
Sources · 1